c038fd4
import * as React from 'react';
import { ConnectionState, ParticipantKind, Track } from 'livekit-client';
import type { RemoteParticipant } from 'livekit-client';
import {
ParticipantAgentAttributes,
type ReceivedTranscriptionSegment,
type TrackReference,
} from '@livekit/components-core';
import { useRemoteParticipants } from './useRemoteParticipants';
import { useParticipantTracks } from './useParticipantTracks';
import { useTrackTranscription } from './useTrackTranscription';
import { useConnectionState } from './useConnectionStatus';
import { useParticipantAttributes } from './useParticipantAttributes';
import { type AgentState } from './useAgent';
/**
* @beta
*/
export interface VoiceAssistant {
/**
* The agent participant.
*/
agent: RemoteParticipant | undefined;
/**
* The current state of the agent.
*/
state: AgentState;
/**
* The microphone track published by the agent or associated avatar worker (if any).
*/
audioTrack: TrackReference | undefined;
/**
* The camera track published by the agent or associated avatar worker (if any).
*/
videoTrack: TrackReference | undefined;
/**
* The transcriptions of the agent's microphone track (if any).
*/
agentTranscriptions: ReceivedTranscriptionSegment[];
/**
* The agent's participant attributes.
*/
agentAttributes: RemoteParticipant['attributes'] | undefined;
}
const state_attribute = ParticipantAgentAttributes.AgentState;
/**
* This hook looks for the first agent-participant in the room.
* @remarks This hook requires an agent running with livekit-agents \>= 0.9.0
* @example
* ```tsx
* const { state, audioTrack, agentTranscriptions, agentAttributes } = useVoiceAssistant();
* ```
* @beta
*/
export function useVoiceAssistant(): VoiceAssistant {
const remoteParticipants = useRemoteParticipants();
const agent = remoteParticipants.find(
(p) =>
p.kind === ParticipantKind.AGENT &&
!(ParticipantAgentAttributes.PublishOnBehalf in p.attributes),
);
const worker = remoteParticipants.find(
(p) =>
p.kind === ParticipantKind.AGENT &&
p.attributes[ParticipantAgentAttributes.PublishOnBehalf] === agent?.identity,
);
const agentTracks = useParticipantTracks(
[Track.Source.Microphone, Track.Source.Camera],
agent?.identity,
);
const workerTracks = useParticipantTracks(
[Track.Source.Microphone, Track.Source.Camera],
worker?.identity,
);
const audioTrack =
agentTracks.find((t) => t.source === Track.Source.Microphone) ??
workerTracks.find((t) => t.source === Track.Source.Microphone);
const videoTrack =
agentTracks.find((t) => t.source === Track.Source.Camera) ??
workerTracks.find((t) => t.source === Track.Source.Camera);
const { segments: agentTranscriptions } = useTrackTranscription(audioTrack);
const connectionState = useConnectionState();
const { attributes } = useParticipantAttributes({ participant: agent });
const state: AgentState = React.useMemo(() => {
if (connectionState === ConnectionState.Disconnected) {
return 'disconnected';
} else if (
connectionState === ConnectionState.Connecting ||
!agent ||
!attributes?.[state_attribute]
) {
return 'connecting';
} else {
return attributes[state_attribute] as AgentState;
}
}, [attributes, agent, connectionState]);
return {
agent,
state,
audioTrack,
videoTrack,
agentTranscriptions,
agentAttributes: attributes,
};
}